The Generation That Grew Up Consuming AI Is Turning Against It. What That Means for Your 9-Year-Old

What this is: a look at new Pew data showing rising AI anxiety among young adults, and a framing question for parents of younger kids — not a claim that any particular product fixes the trend. We don't have data showing that building with AI changes how a kid feels about it later. We have a question worth asking before you hand a 9-year-old their first AI tool.

The data

Pew's 2026 Global Attitudes Survey (36 countries, n=42,151, fielded February to May 2026) found something specific and a little uncomfortable: the people most likely to say they're "more concerned than excited" about AI's growing role in daily life are young adults — the demographic that supposedly grew up fluent in this stuff.

In the US, 18-to-34-year-olds went from 40% "more concerned than excited" in 2024 to 55% in 2026. Sweden moved from 24% to 45%. Pew also found statistically significant rises in Poland, Brazil, Japan, and Australia. Young adults in the survey were more worried than older adults about AI costing jobs, and more likely to report feeling both excited and concerned at the same time — not simple optimists, not simple skeptics, just uneasy.

This is the age group that met ChatGPT, TikTok's algorithm, and Instagram's AI filters as fully built, polished products. They never saw the scaffolding. They just got the output — and now, a couple of years in, a majority of them say the feeling is closer to dread than enthusiasm.

We want to be precise about what this data does and doesn't say. Pew didn't ask why concern is rising, and didn't test whether a different first encounter with AI — one where you built something instead of just consuming something — would change the trajectory. Nobody has run that study. We're not going to pretend we have.

What we think is going on

Here's a theory, and it's a theory, not a finding: this generation's first relationship with AI was almost entirely as a consumer. They scrolled a feed an algorithm built for them. They asked a chatbot to write the essay, or summarize the reading, or finish the email. The AI did something to them or for them, long before most of them ever directed it to do something they specifically wanted.

The OECD's 2026 Digital Education Outlook makes a related point in an education context: when students use general-purpose generative AI to complete tasks without any pedagogical design around it, they outsource the work. Output quality goes up. Learning doesn't follow. In a field experiment in Türkiye high schools that OECD cites, students with GenAI access scored about 127% higher on practice problems — and about 17% worse on exams once the AI access was removed. The advantage was borrowed, not built.

That's not a study about emotional attitudes toward AI. But it rhymes with the Pew finding. If your first few years with a technology are mostly about handing it tasks and watching it perform, you never develop a felt sense of how it works, where it fails, or what you're actually in control of. That's a reasonable setup for anxiety once the technology starts feeling bigger than you.

One more data point, from a different source: on a September 2026 Edtech Insiders webinar, Dan Meyer of Amplify (formerly of Desmos) said something worth sitting with — that kids who have AI access at home mostly use it to get their homework done for them, not to learn anything from it. He wasn't talking about Pew's cohort; he was talking about kids right now. But it's the same pattern one generation earlier: AI as something that produces an output for you, not something you direct.

The parent question

So here's the actual question, stripped of product pitches: when your kid meets generative AI for the first time, do you want that first relationship to be as a consumer of AI output, or as a director of it?

A consumer relationship looks like: the AI writes the essay, finishes the homework, generates the image, picks the next video. The kid's job is to receive and evaluate — sometimes not even that.

A director relationship looks like: the kid describes what they want, gets something back that's wrong or boring or not quite it, and has to say specifically what to change. "The dragons need to be faster." "Make the quiz harder." "That's not the joke I meant, try again." The kid is making judgment calls about someone else's (something else's) output, over and over, instead of just accepting the first draft.

We can't tell you that the second relationship produces an adult who's less anxious about AI at 25. We don't have that data, and anyone who tells you they do is making it up. What we can tell you is that it's a different relationship, started earlier, and it's worth having instead of defaulting into the first one by accident — which is what happens if a kid's introduction to AI is a chatbot that does their homework or a feed that picks what they watch next.

Where the caution should actually live

This isn't an argument for handing a young kid unrestricted AI access. Common Sense Media's age-and-stage guidance (published June 2026) recommends no AI chatbots at all for kids 5 and under — young children genuinely can't tell AI apart from a real person — and for ages 6 to 12, "limited use, alongside a parent or caregiver." Their guidance also offers a useful diagnostic question for any AI tool, chatbot or otherwise: is this AI helping me learn, or just helping me finish?

That question applies to creative tools too, not just chatbots, and it's worth asking honestly. A tool where a kid types one prompt and gets a finished, polished game is helping them finish. A tool where the first result is good but not quite right, and the kid has to say exactly what's wrong and try again, is closer to helping them learn — in the narrow, specific sense of learning to direct the thing, not in the sense of learning math or reading.

We should also be honest that institutions are moving the other direction right now, for good reasons. NYC's Department of Education put a one-year moratorium on student-facing generative AI for grades K through 8 in September 2026, covering roughly 600,000 students, alongside recommended daily screen-time caps (30 minutes for grades 3–5, 45 minutes for grades 6–8). LAUSD blocked generative AI on all district student devices for all ages the same month, across roughly 378,000 students, with recommendations due by the end of the 2026–27 school year. Those are reasonable, cautious responses to a technology schools don't yet have guardrails for in a classroom setting. They're about school deployment at scale — they're not a verdict on a parent, at home, deciding how their own kid meets AI for the first time with full visibility into what's happening.

What this looks like in practice

At Xyplor, we built our own small, honest signal into this question rather than guessing. Among the first 71 kid revisions we tagged internally, 55% were the kid rejecting what the AI built and asking for something different — not adding to it, not restarting from scratch, actively pushing back on a first draft. That's an internal tag count from a small sample, not a study, and it will move as volume grows. We're not claiming it proves kids who build develop better judgment about AI, or less anxiety about it later. We're saying: when a kid's workflow requires telling an AI "no, that's wrong, try this instead" more than half the time, the kid is doing something structurally different than accepting output.

Whether that difference changes how a 9-year-old feels about AI at 25 is a question nobody has answered yet, including us. It'll take a decade of data that doesn't exist. What we'd rather do today is be honest about the shape of the question, point you to Common Sense Media's age guidance, and let you make the call for your own kid with full information rather than a growth-hack promise.

If the choice in front of you is "my kid is going to encounter generative AI somehow, soon" — and for most families with a kid over six, that choice has already arrived — the framing we'd offer is simple: notice whether the first version of AI your kid meets asks anything of them, or just performs for them. That's the whole argument. We don't know yet whether it changes the Pew numbers in fifteen years. We know it's a different way to start.